About the Topic
Product and process complexity in semiconductor manufacturing continues to rise steadily. However, even highly automated fabs like Infineon suffer from adverse and growing variability and losses in production effectiveness. This leads to unscheduled and costly delays in deliveries to customers. One reason is a growing gap regarding adequate information modeling of the manufacturing process – even in highly automated fabs.
Bayesian inference is a method for analyzing and understanding causal relationships. This method has gained a lot of importance in recent years - especially with respect to machine learning. The focus of this work is to explore the applicability of Bayesian inference to problems in semiconductor manufacturing. This is supported and enabled by a new, innovative information model, called the holistic information model (HIM).1 HIM provides, for the first time, all the necessary data structures to directly explore and apply Bayesian inference. This may (but need not) include the use of new machine learning methods. SYSTEMA will provide a complete software framework, including the holistic information model HIM. Sample production data is available.
More information on the current research together with Infineon.
One outcome of the work is a proposal for new methods to improve productivity of production lines and overcome the problems mentioned above. Candidates should have a background and interest in one of the following areas: Computer Science, Mathematics or Operations Research.